Got it to work like so: read(from.textFile(inputPath)).write(to.textFile(outputPath)).native.getPipeline().done()
Is that the correct way? Thanks for the help, I have a running word count example now. :-) On Fri, Jun 20, 2014 at 4:34 PM, Josh Wills <[email protected]> wrote: > You need to manually call run() or done() to execute the pipeline if > you're not materializing the output. The user guide will be useful for the > basic concepts, even though it focuses on the Java API. > On Jun 20, 2014 1:27 PM, "Daniel Siegmann" <[email protected]> > wrote: > >> Thanks Josh! The thrift and protobuf defs were what I was missing. I'm >> able to compile and run the code now. I also updated to Scrunch 0.10.0. >> >> Any idea why it might not write the output? If I have >> >> countWords(args(0)).materialize.foreach(line => println(s"**** $line")) >> >> I get all my output, but >> >> countWords(args(0)).write(to.textFile(args(1))) >> >> Doesn't even create the output directory, even though I see this in my >> logs >> >> 14/06/20 16:17:47 INFO impl.FileTargetImpl: Will write output files to >> new path: >> /var/folders/th/7vf9rjqd1955jnwnzg3x9ym40000gn/T/1403295466563-1/wordcounts >> >> No exceptions or anything. I'm probably missing something obvious. :-( >> >> >> On Thu, Jun 19, 2014 at 6:03 PM, Josh Wills <[email protected]> wrote: >> >>> Here you go: https://github.com/jwills/scrunch-demo >>> >>> Did this w/Maven; you'll have to forgive me as my SBT-fu isn't great. It >>> looks like vanilla Hadoop 1.x doesn't include any thrift/protobuf >>> dependencies that Scrunch expects to be present at compile-time; I added >>> them as provided dependencies in this example and then verified that I >>> could run the -job.jar that I built w/mvn package under Hadoop 1.0.3. >>> >>> J >>> >>> >>> On Thu, Jun 19, 2014 at 2:33 PM, Daniel Siegmann < >>> [email protected]> wrote: >>> >>>> Hi Josh, thanks for the reply. >>>> >>>> Which version of Hadoop are you looking to compile against? >>>>> >>>> >>>> I think any 1.x version will suffice (our production cluster is MapR). >>>> >>>> The Spotify comparison is interesting. Too bad they didn't evaluate >>>> Scoobi as well. Thanks for the info. >>>> >>> >>> >>> >>> -- >>> Director of Data Science >>> Cloudera <http://www.cloudera.com> >>> Twitter: @josh_wills <http://twitter.com/josh_wills> >>> >> >> >> >> -- >> Daniel Siegmann, Software Developer >> Velos >> Accelerating Machine Learning >> >> 440 NINTH AVENUE, 11TH FLOOR, NEW YORK, NY 10001 >> E: [email protected] W: www.velos.io >> > -- Daniel Siegmann, Software Developer Velos Accelerating Machine Learning 440 NINTH AVENUE, 11TH FLOOR, NEW YORK, NY 10001 E: [email protected] W: www.velos.io
